April 2007doc.: IEEE 802.22-07/0203r0
IEEE P802.22
Wireless RANs
Date: 2007-04-18
Author(s):
Name / Company / Address / Phone / email
Gerald Chouinard / Communications Research Centre, Canada / 3701 Carling Ave.Ottawa, OntarioCanadaK2H 8S2 / 613-998-2500 /
Andre Brind’Amour / Communications Research Centre, Canada / 3701 Carling Ave.Ottawa, OntarioCanadaK2H 8S2 / 613-998-2853 /
Hisham Hassanein / Communications Research Centre, Canada / 3701 Carling Ave.Ottawa, OntarioCanadaK2H 8S2 / 613-998-2462 /
Submissionpage 1Gerald Chouinard, CRC
April 2007doc.: IEEE 802.22-07/0203r0
VoIP bit rate requirement
This contribution summarizes the key parameters of various audio codecs used for telephony in wired and wireless communications networks and describes the structure of the information needed to establish a VoIP communication. The purpose is to determine the bit rate requirement for this real-time application in order to assist in the determination of the minimum data block size to be carried in a WRAN transmission frame.
- Narrowband codecs
These codecs provide for a speech bandwidth of less than 4 kHz.
Codec / Organi-zation / Date / Type / System use / Bit rate (kbit/s) / Delay frame size (msec) / Delay lookahead (msec) / Quality / Complexity MIPS / Complexity RAM (byte)G.711 / ITU / 1972 / Companded PCM / PSTN, Packet Network / 64 / 0.125 / 0 / Toll / <1 / 1
G.726-G.727 / ITU / 1984, 1986 / ADPCM / PSTN, Packet Network / 16, 24, 32, 40 / 0.125 / 0 / Toll / 1.25 / <50
G.728 / ITU / 1992, 1994 / LD-CELP / Packet Network / 16 / 0.625 / 0 / Toll / 30-40 / 2K
G.723.1 / ITU / 1995 / MP-MLQ/
ACELP® / Packet Network / 6.3/
5.3 / 30 / 7.5 / <Toll / 20-25 / <3K
G.729 / ITU / 1995 / CS-ACELP® / Packet Network, Wireless / 8 / 10 / 5 / Toll / 20-25 / <4K
G.729a / ITU / 1996 / CS-ACELP® / Packet Network (DSVD) / 8 / 10 / 5 / Toll / 10 / 2K
G.729d / ITU / 1998 / CS-ACELP® / Packet Network / 6.4 / 10 / 5 / <Toll / 20 / <4K
G.729e / ITU / 1998 / CS-ACELP® / Packet Network (music +) / 11.8 / 10 / 5 / <Toll / 25-30 / <6K
Full Rate GSM / ETSI / 1987 / RPE-LTP / Cellular / 13 / 20 / 0 / <Toll / 4.5 / 1K
Half Rate GSM / ETSI / 1994 / VSELP / Cellular / 5.6 / 20 / 5 / <Toll / 30 / 4K
EFR GSM / ETSI / 1996 / ACELP® / Cellular / 12.2 / 20 / 0 / Toll / 15-20 / 4K
AMR-NB / ETSI / 1999 / ACELP® / Wireless Multimedia/3G / 4.75, 5.15, 5.9, 6.7, 7.4, 7.95, 10.2, 12.2 / 20 / 5 / Toll for higher rates
(link controlled) / 15-20 / 4K
Tetra / ETSI / 1994 / ACELP® / Cellular / 4.63 / 30 / 5 / <Toll / 10-15 / 4K
IS-54 / TIA / 1989 / VSELP / Cellular (TDMA) / 7.95 / 20 / 5 / <Toll / 20 / 2K
IS-96 / TIA / 1993 / QCELP / Cellular (CDMA) / 0.8, 2, 4, 8.5 / 20 / 5 / <Toll
(source controlled) (<IS-54) / 20 / 2K
IS-127 (EVRC) / TIA / 1997 / RCELP / Cellular (EVRC/
CDMA) / 0.8, 2, 4, 8.5 / 20 / 5 / Toll
(source controlled) / 25-30 / 4K
IS-733 (Q-13) / TIA / 1995 / CELP / Wireless (CDMA) / 1.0, 2.7, 6.2, 13.3 / 20 / 5 / Toll
(source controlled) / 20-25 / 4K
IS-641 (IS-136) / TIA / 1995 / ACELP® / Wireless (TDMA) / 7.4 / 20 / 5 / Toll / 15 / 4K
PDC
Full Rate / ARIB / 1990 / VSELP / Wireless (TDMA) / 6.7 / 20 / 8 / <Toll
(<IS-54) / 12
PDC Half Rate / ARIB / 1993 / PSI-CELP / Wireless (TDMA) / 3.45 / 40 / 5 / <Toll / 25
PDC EFR / ARIB / ACELP® / Wireless (TDMA) / 6.7 / 20 / 5 / <Toll / 15
2. Wideband codecs
These codec extend the speech bandwidth to 50 Hz - 7 kHz to offer a communication quality surpassing that of PSTN and give a sensation of face-to-face communication quality. The low frequency extension contributes to increased naturalness, presence and comfort while the high frequency extension provides for higher intelligibility. Such higher level service is available as long as wideband codecs are available at both ends of the communication and there is enough bandwidth to support the higher bit rates. If the communication has to be bridged to the PSTN, the extended bandwidth will not be available.
Codec / Date / Type / Bit rate (kbit/s) / Delay frame size (msec) / Delay lookahead (msec) / Quality / Complexity MIPS / Complexity RAM (byte)G.722 / 1988 / Sub-band ADPCM / 64, 56, 48 embedded / 0.125 / 1.5 / Commentary at 64 kbit/s / 10 / 1Kwords
G.722.1 / 1999 / Transform Coding / 32, 24 / 20 / 20 / Hands free and low packet loss rate
Good music performance / <15 / 2Kwords
G.722.2 / 2001/2002 / ACELP® / 23.85, 23.05, 19.85, 15.85, 14.25, 12.65, 8.85, 6.6 / 20 / 5 / Good speech performance at >12.65 kbit/s / 38 WMOPS / 5.3 KB
Submissionpage 1Gerald Chouinard, CRC
April 2007doc.: IEEE 802.22-07/0203r0
3. VoIP bit rate requirement
Extra header information is required to establish the real time operation of voice channels over Internet protocol as shown in the following figure. Here is an example for the G.729 codec operating at 8 kbit/s with a frame size of 10 ms, which means that the packet below needs to be sent every WRAN frame. This packet represents a data block size of 56 bytes or 448 bits to be transmitted every frames or the equivalent of a 44.8 kbit/s bit stream. Note that if this IP packet were to also include the codec payload from the previous VoIP packet to improve reliability against packet loss, the packet length would increase by only 18%.
MAC header6 bytes / IPv4 header
20 bytes / UDP header
8 bytes / RTP header
12 bytes / Voice codec payload
10 bytes
3. VoIP latency budget
A well accepted end-to-end latency for telephony is 150 ms. If we assume that the total codec delay for the G.729 implementation is 25 ms (10 ms delay frame size, 5 ms delay look ahead and 10 ms processing delay) and that the decoding at the other end represent a negligeable delay, the remaining allowable delay for the transmission is 125 ms. Depending on the delay/jitter tolerance for the backhaul data network, a certain delay could be budgeted for the WRAN transmission. At the edge of the contour, i.e., 30 km, the RF path delay is 0.1 ms which is negligeable in this context but the transmission delay due to frame construction is 10 ms per frame. For example, if the backhaul data network is allowed to create a 84 ms latency/jitter, then the delay budget can be as follows:
Input codec:25 ms
WRAN framing:20 ms
Backhaul latency:84 ms
WRAN framing:20 ms
Output codec: 1 ms
Total:150 ms
In this case, the WRAN networks would be allowed to concatenate two codec frames in every other WRAN frame, making it more efficient and easier in terms of QoS.
4. Packet header compression
In order to reduce the sizeable overhead brough about by VoIP, work is being done in various organizations such as the 3GPP to implement header compression. More work is needed to document the extent of such header compression mechanisms. Such header compression could bring the IP header size from 46 bytes, as seen above, down to 10 bytes, which would result in datablock sizes as indicated in the following table.
Audio frame grouping (frame) / Datablock size (bytes) / Datablock size (bits) / Data rate (kbit/s) / WRAN framing delay (msec)1 / 10+10 / 160 / 16 / 10
2 / 10+20 / 240 / 24 / 20
3 / 10+30 / 320 / 32 / 30
Note that the granularity of a subchannel for a 7 symbol upstream burst is 16.8 kbit/s when the OFDMA multiplex contains 60 sub-channels.
5. VoIP session signalling
The signalling to establish the VoIP session (e.g., dialing & ringing and maintaining the session) will also need to be sent over the transmission channel in different packets. These extra packets will be needed at the start, during and the end of the VoIP session and will occupy less transmission capacity than the streaming packets between the codecs. This signalling will most likely follow the session initiation protocol (SIP) as defined by the IETF (also H223, ITU). Other side packets will also likely be required to maintain the real-time transmission (RTCP) but the capacity requirement will be minimal.
6. Reference
Jeff Tyre and Roger Britt, “Performance Characteristics of Voice over IP Networks”, Nortel Application Note ANQOSIPT03 v1.0,
______
Submissionpage 1Gerald Chouinard, CRC